I want to make json with the help of page.extract_words and im.draw_rects

[Faizan-Saifullah]

Hi
I have extracted words from a pdf page and i have been able to draw rectangles using im.draw_rects on those extracted words. Now i want to make json file that contains all this information of page.extract_words such that when i open in labelme , it shows same rects as im.draw_rects .

Here , is my code:

with pdfplumber.open(r'empty_test.pdf') as pdf:

print(pdf.pages)
page = pdf.pages[0]
im = page.to_image(resolution = 400)
text = page.extract_words()
im = im.draw_rects(page.extract_words())
im.save("result.jpeg")
#h = open('empty_test' + '.json', "w")
#json.dump(text, h, indent=2, sort_keys=False)
#h.close()

imm = im.annotated
imm.save("result2.jpeg")
print(page.extract_words())
ann = im.original
ann.save("scan.jpeg")

[samkit-jain]

Hi @Faizan-Saifullah Appreciate your interest in the library. To help you further, could you please share the JSON structure format in which you want to store the data?

[Faizan-Saifullah]

the structure is almost the same as the one being obtained from "print(page.extract_words())" :

[{'text': 'Requisition', 'x0': Decimal('217.800'), 'x1': Decimal('251.814'), 'top': Decimal('10.440'), 'bottom': Decimal('17.440'), 'upright': True, 'direction': 1}, {'text': .....

Such that when I open in labelme , it displays the bounding boxes as well as extracted text.

[Faizan-Saifullah]

Above is Done!
Just one more thing (if you can please):
I am unable to convert bbox coordinates from pdf coords to image coords, so that the bbox for an "extracted_word" is displayed at correct position in the following file created in code: "ann.save("scan.jpeg")"

[samkit-jain]

You can use the protected method PageImage._reproject_bbox(...) to do that. It will give you the coordinates translated to the image. Example usage,

import pdfplumber

pdf = pdfplumber.open("file.pdf")
page = pdf.pages[0]
image = page.to_image(resolution=200)

for word in page.extract_words():
    x0, top, x1, bottom = image._reproject_bbox((word["x0"], word["top"], word["x1"], word["bottom"]))